SUMMARYPneumonia and other lower respiratory tract infections are the leading contributors to global mortality of any communicable disease [1]. During normal pulmonary homeostasis, competing microbial immigration and elimination produce a transient microbiome with distinct microbial states [2–4]. Disruption of underlying ecological forces, like aspiration rate and immune tone, are hypothesized to drive microbiome dysbiosis and pneumonia progression [5–7]. However, the precise microbiome transitions that accompany clinical outcomes in severe pneumonia are unknown. Here, we leverage our unique systematic and serial bronchoscopic sampling to combine quantitative PCR and culture for bacterial biomass with 16S rRNA gene amplicon, shotgun metagenomic, and transcriptomic sequencing in patients with suspected pneumonia to distill microbial signatures of clinical outcome. These data support the presence of four distinct microbiota states—oral-like, skin-like,Staphylococcus-predominant, and mixed—each differentially associated with pneumonia subtype and responses to pneumonia therapy. Infection-specific dysbiosis, quantified relative to non-pneumonia patients, associates with bacterial biomass and elevated oral-associated microbiota. Time series analysis suggests that microbiome shifts from baseline are greater with successful pneumonia therapy, following distinct trajectories dependent on the pneumonia subtype. In summary, our results highlight the dynamic nature of the lung microbiome as it progresses through community assemblages that parallel patient prognosis. Application of a microbial ecology framework to study lower respiratory tract infections enables contextualization of the microbiome composition and gene content within clinical phenotypes. Further unveiling the ecological dynamics of the lung microbial ecosystem provides critical insights for future work toward improving pneumonia therapy.